You are here

Harvard Forest >

Harvard Forest REU Symposium Abstract 2010

  • Title: Addressing uncertainties associated with computational models, to improve the accuracy of forecasting species distribution response to climatic change
  • Author: Elisabete M Baker (Simmons College)
  • Abstract:

    Principle Investigator – Aaron Ellison, Matthew Fitzpatrick

    REU Advisor- Sydne Record

    REU Student- Elisabete Vail (formerly Baker)



    Ecology has recently begun to depend heavily on the use of computational models to make predictions of how climate change will affect our planet. However, unaccounted “uncertainties” caused by inherent variation or often complete omission of projected data, have lead to disparity among current model forecasts, casting doubt on their presumed reliability. The greater project goal strives to improve the accuracy of these predictions, by evaluating relevant computer models and ultimately developing a new model that will address uncertainties currently overlooked by others. With a focus on the impact of how climate/species relationships will influence species distribution across both spatial and temporal scales - my role in this project was to create three test datasets, representing one present and two future climate scenarios of species specific climatic variables for 27 local Oak species. Each dataset contains the top five important variables for a given species, selected from 23 initial categories; using averaged results produced from 500 iterations of 1000 Classification and Regression Trees (CART). Species presence/absence data from over 50,000 Forest Inventory and Analysis (FIA) plots spanning the eastern United States were also included. Uncertainties sometimes caused by model type, were compensated for by analyzing the datasets using an ensemble forecasting approach of nine models featured in the R based statistical software package – BIOMOD. I will present the process involved in creating the datasets along with the results for two oak species, Quercus rubra (red oak) and Quercus virginiana (water oak). Although, these results were produced as reference data for the larger model development project – they yield interesting implications regarding the potential affect climate change may have on the distribution of important foundation species, such as oak, while further demonstrating the power models possess and the motivation behind investing effort into improving their accuracy.

  • Research Category: Ecological Informatics and Modelling